Anthropic Claude Faces Large-Scale Distillation Attacks by Chinese AI Labs Involving 13 Million Exchanges

Anthropic has revealed that three Chinese AI laboratories DeepSeek, Moonshot AI, and MiniMax, conducted industrial-scale distillation campaigns targeting its Claude AI models.

The three labs collectively generated over 16 million exchanges with Claude through approximately 24,000 fraudulent accounts, violating Anthropic’s terms of service and regional access restrictions.

Distillation is a machine learning technique where a less capable AI model is trained on the outputs of a more advanced one.

While this method is commonly and legitimately used by AI labs to create smaller, cheaper versions of their own models, it can also be weaponized by competitors.

In this case, foreign labs used distillation to extract Claude’s most powerful capabilities: agentic reasoning, tool use, and coding at a fraction of the time and cost it would take to develop them independently.

These campaigns followed a similar playbook. The labs used fraudulent accounts and commercial proxy services to access Claude at scale while evading detection.

https://twitter.com/AnthropicAI/status/2025997928242811253?ref_src=twsrc%5Etfw

Anthropic attributed each campaign with high confidence through IP address correlation, request metadata, infrastructure indicators, and corroboration from industry partners.

Scale of the Attacks

MiniMax was responsible for the largest operation, generating over 13 million exchanges focused on agentic coding and tool use.

Anthropic detected this campaign while it was still active before MiniMax released the model it was training.

When Anthropic released a new Claude model during the campaign, MiniMax pivoted within 24 hours, redirecting nearly half its traffic to capture capabilities from the latest system.

Moonshot AI (Kimi models) conducted over 3.4 million exchanges targeting agentic reasoning, coding, computer-use agent development, and computer vision.

The lab employed hundreds of fraudulent accounts across multiple access pathways, and request metadata matched the public profiles of senior Moonshot staff.

DeepSeek, the smallest campaign at over 150,000 exchanges, focused on reasoning capabilities and rubric-based grading tasks.

In a notable technique, DeepSeek prompted Claude to articulate internal reasoning behind completed responses step by step, effectively generating chain-of-thought training data at scale.

DeepSeek also used Claude to generate censorship-safe alternatives to politically sensitive queries about dissidents, party leaders, and authoritarianism.

Lab	Exchanges	Primary Targets
MiniMax	Over 13 million​	Agentic reasoning, coding,and computer vision
Moonshot AI	Over 3.4 million​	Agentic reasoning, coding, computer vision
DeepSeek	Over 150,000​	Reasoning, chain-of-thought extraction, censorship-safe content

In one case, a single proxy network managed more than 20,000 fraudulent accounts simultaneously, mixing distillation traffic with unrelated customer requests to make detection harder.

OpenAI has also accused DeepSeek of using similar distillation techniques against ChatGPT.

According to a memo sent to U.S. lawmakers, DeepSeek employees used third-party routers and unauthorized resellers to circumvent OpenAI’s access restrictions.

Anthropic warns that illicitly distilled models lack the safety guardrails that U.S. companies build into their systems.

Anthropic has deployed multiple countermeasures, including classifiers and behavioral fingerprinting systems to detect distillation patterns in API traffic, chain-of-thought elicitation detection, and tools for identifying coordinated multi-account activity.

Anthropic is sharing technical indicators with other AI labs, cloud providers, and relevant authorities, and is developing model-level safeguards designed to reduce the effectiveness of outputs for illicit distillation without degrading the legitimate user experience.

Anthropic stressed that no single company can solve this problem alone, calling for coordinated action across the AI industry, cloud providers, and policymakers to address distillation threats at this scale

Follow us on Google News, LinkedIn and X to Get More Instant Updates. Set Cyberpress as a Preferred Source in Google

The post Anthropic Claude Faces Large-Scale Distillation Attacks by Chinese AI Labs Involving 13 Million Exchanges appeared first on Cyber Security News.